Optical Coherence Tomography (OCT) is a novel imaging technique which allows for noninvasive cross-sectional imaging in scattering or cloudy media with high spatial resolution and high dynamic range. OCT is a two-dimensional extension of Optical Coherence-Domain Reflectometry (OCDR) which is also commonly referred to as Optical Low Coherence Reflectometry (OLCR), in which a low temporal coherence light source is employed to obtain precise localization of reflections internal to a probed structure along the optic axis. The one-dimensional ranging technique of OCDR/OLCR has previously been utilized for characterization of bulk-, integrated-, and fiber-optic structures, as well as biological tissues. In OCT, this technique is extended to provide for scanning of the probe beam in a direction perpendicular to the optic axis, building up a two-dimensional data set comprising a cross-sectional image of internal tissue backscatter.
Ophthalmic Applications of OCT
OCT has previously been applied to imaging of biological tissues in vivo and in vitro, although the majority of initial biomedical imaging studies concentrated on transparent structures such as the eye. Initial ophthalmic imaging studies demonstrated significant potential for OCT imaging in routine examination of normal and abnormal ocular structures, including imaging of the cornea, iris, and other structures of the anterior eye; the lens and lens capsule; and numerous structures in the posterior eye, including the neurosensory retina, retinal nerve fiber layer, retinal pigment epithelium, and choroid. In OCT examination of the retina, initial in vivo clinical studies have demonstrated its utility in aiding diagnosis in a variety of vitreoretinal diseases, including macular hole, macular degeneration, detached retina, and glaucoma. Clinical trials of OCT imaging for ophthalmic applications are currently under way at several centers, and a commercial ophthalmic OCT scanner is available from Humphrey Systems of Dublin, Calif.
OCT Imaging in Highly Scattering Media
Several recent publications have demonstrated the potential applications of OCT in highly scattering media for the measurement of tissue optical properties and imaging. Optical imaging in scattering media such as biological tissue is in general a very difficult problem, particularly for techniques such as OCT which depend primarily upon unscattered or singly-scattered light for image formation. It has been observed in preliminary studies and theoretical treatments that this singly-scattered gating requirement practically limits OCT imaging to a useful penetration depth of a few millimeters at best in nontransparent human tissues. Nonetheless, several authors have identified diagnostic scenarios in which a technique for improved, non-invasive 10–20-micron scale optical imaging near tissue surfaces has significant potential for clinical utility. These include applications of OCT imaging in skin, teeth, vascular tissues, and gastrointestinal mucosa. The latter two examples are significant since with its fiber optic implementation, OCT is readily adaptable to minimally invasive diagnostic modalities such as catheterization or endoscopy. OCT system implementations featuring the high-speed imaging acquisition necessary for in vivo application and catheter/endoscopic delivery have been reported. The application of OCT to biomedical imaging provides the potential for sub-surface tissue characterization with sufficient resolution to provide microscopic morphological information relevant to pathological diagnosis without the need for biopsy.
OCT Imaging in Industrial Processing
Recent publications have also illustrated the potential applications of OCT for imaging in cloudy or turbid non-biological media in industrial processing in the manufacturing industry. OCT imaging may be useful for on-line process control or product testing and evaluation. Initial experiments have demonstrated OCT imaging in ceramic and other highly scattering materials, as well as for the characterization of the surface topology of opaque industrial materials such as metals (i.e., ball bearings).
OCT Qualitative Technical Description
Optical coherence tomography performs micron-scale topographic imaging of internal tissue microstructure using a combination of the principles of low-coherence interferometry and confocal microscopy. Reference is made to FIG. 1 illustrating an exemplary OCT system 10 in which the tissue to be examined is placed in the sample arm 12 of a Michelson interferometer illuminated by a broadband light source 16. Due to the limited coherence length of the source (typically 10–15 microns), light returning from the reference arm 18 and light backscattered by internal sample reflections interferes constructively or destructively only when the interferometer arm optical path lengths are matched to within the source coherence length. Scanning the reference arm 18 length through a position corresponding to the depth of a reflecting site within the sample generates a localized interference pattern, which is recorded as a localized modulation of the detector current as a function of the reference arm position. A beamsplitter 20, optical detector 22, transimpedance amplifier 24, demodulator 26, A/D converter 28, and display 30 are also shown. The detector current generated by a sample containing multiple reflecting sites distributed along its depth (such as biological tissue) contains the sum of multiple, overlapping copies of this interference pattern. A map of tissue reflectivity versus depth or “A-Scan” is obtained by scanning the reference mirror 32 at constant velocity, while recording the envelope of the detector current. The envelope may be recorded with high dynamic range by scanning the reference mirror 32 at fixed velocity, and demodulating the detector current at the resulting Doppler frequency. Cross-sectional images of tissue backscatter or “B-Scans” may be acquired by obtaining sequential A-scans while scanning the probe beam across the tissue surface using a lateral beam scanning mirror 33 or some other lateral scanning optic device. The resulting two-dimensional datasets are plotted as gray-scale or false-color images.
A significant advantage of using low-coherence interferometry for signal detection is that the interferometer 14 acts as an optical heterodyne detector, providing a dramatic expansion in dynamic range compared to direct detection of scattered light. Since the interferometric component of the detector current is proportional to the product of the electric field amplitudes returning from each arm, the detected envelope signal is proportional to the square root of the sample power reflectivity. Extremely faint reflections in the sample (˜10−11 times the incident optical power) are routinely detected in A-scans recorded in a fraction of a second. As illustrated in FIG. 1, the interferometer 14 can also be implemented using inexpensive semiconductor sources and detectors, and flexible single-mode optical fibers suitable for remote imaging through minimally invasive diagnostic instruments.
Signal-To-Noise Ratio in OCDR and OCT
A significant limitation in the use of OCDR and OCT in highly scattering media is that the OCT probe light is very strongly (exponentially) attenuated in the scattering material, thus limiting the imaging depth which can be achieved in a given amount of time for a given sensitivity. For a conventional OCT system in which a 50/50 beamsplitter 20 is used in the Michelson interferometer, the signal to noise ratio (SNR) of the detected electronic signal in the shot-noise limit is given by Eq. (1) below:
                    SNR        =                              ρ            ⁢                                                  ⁢                          P              s                        ⁢                          R              s                                            2            ⁢            qB                                              (        1        )            In this expression, SNR is signal-to-noise ratio (a measure of the sensitivity which also relates to imaging depth in scattering media), ρ is the detector responsivity, Ps is the optical power incident on the sample, Rs is the optical power reflectivity of the sample, q is the charge on the electron, and B is the detector bandwidth. The latter variable B is inversely proportional to the time required to obtain an OCDR scan or OCT image. The shot-noise limit under which this expression is calculated is well known to those practiced in the art to be the best possible performance (i.e., to give the best value for S/N) which can be achieved in an optical detection system. Even though not all implementations of OCDR aid OCT may actually achieve true shot-noise limited performance and therefore may not be strictly governed by Eq. (1), most implementations aim to be near this limit and the equation is still a useful guideline illustrating the trade-offs between sensitivity, source power, and image acquisition time in this limiting case of the best possible performance.
Equation (1) makes clear that there is a trade-off between sensitivity or depth, imaging time, and the source power incident on the sample in OCDR and OCT. Increases in imaging speed, for example, may only be achieved through either a decrease in S/N or an increase in power incident on the sample. Increases in sensitivity or imaging depth (both proportional to S/N) may only be obtained by increasing either the imaging time or the power on the sample. For industrial and medical imaging applications, it is desirable to image as rapidly as possible, at a rate of at least several images per second. Recently, new technology has been developed permitting OCT image acquisition up to video rate (30 images/second), and high power low-coherence sources have become available to partially compensate for the decrease in sensitivity which necessarily accompanies any increase in imaging speed according to Eq. (1). However, these high power sources are very expensive, and still are not sufficiently powerful to allow for clinically acceptable quality imaging at video rate (or even at the ˜10 images/second rate common to ultrasound imaging).
Detector Power Limitations for Shot-Noise Limited Performance
Two requirements on the amount of optical power which may be incident on the detector must be met in order to be at or near the shot-noise limit in OCDR and OCT. The first requirement is that the total optical power incident on the detector 22 cannot be arbitrarily high in order for shot noise to dominate over excess intensity noise for available sources. For systems with optical sources 16 which emit low power, this is not a problem. However, recent developments in source technology have resulted in the availability of higher power sources (10–20 mW) which are very attractive for high-speed imaging since the higher sample arm power partially compensates for the increased bandwidth B necessary for higher speed imaging, according to Eq. (1). Since most industrial and biological samples have very low reflectivity, they do not reflect an appreciable amount of sample arm light power onto the detector 22. However, in conventional systems employing such high power sources, an attenuator must be placed in the reference arm 18 in order to approach the shot noise limit. This represents a waste of up to 50% of the valuable and expensive source power, which is lost in an attenuator. It would be much better if this power could instead be directed onto the sample, so it could contribute to imaging performance as describe in Eq. (1). Clearly there is a need for an improved interferometer design for OCDR and OCT which avoids power losses due to attenuation required to achieve shot-noise-limited performance on the detector.
The second requirement on the amount of power on the detector is that it must be sufficiently high so that shot noise dominates over thermal noise in the detector. For most commonly available semiconductor detectors in the visible and near-infrared regions of the spectrum, the power on the detector must be at least approximately 1 μW for a typical low speed system using a detector with a bandwidth less than approximately 100 kHz, to 10 μW for a typical high speed system using a detector with a bandwidth of approximately 10 MHz. Thus, there is a range of acceptable power levels which will achieve shot-noise limited performance at the detector, and under the assumption that most of the light reaching the detector comes from the reference arm (i.e., under the approximation of a weakly reflecting sample), this places a limitation on the range of acceptable power levels in the reference arm, which is typically in the range of between 1 μW and 10 μW.
Reciprocal Optical Elements: The Beamsplitter/Fiber Coupler
In conventional OCDR and OCT, the central element of the Michelson interferometer is a standard beamsplitter 20 which transmits or splits some fraction of the power (typically 50%) of the incident light power into each of the sample and reference arms 12 and 18. In a bulk optic interferometer the beamsplitter 20 may be a mirror with a partially reflective coating, while in a fiber optic interferometer the beamsplitter is composed of a pair of fibers partially fused together which is known as a fiber coupler. As illustrated in FIG. 2, the beamsplitter may be abstracted as a four-port optical element with two inputs (labeled as I1 and I2), and two outputs (labeled as O1 and O2). The abstracted beamsplitter illustrated in FIG. 2 is characterized by a splitting ratio α, such that a fraction α of the light power incident at port I1 (neglecting small internal losses of the beamsplitter) is transmitted to port O2, while the fraction (1−α) of the light power incident at port I1 is transmitted to port O1. A similar statement applies to light power incident at port I2: in this case, a fraction α of the light power incident at port I2 (neglecting small internal losses of the beamsplitter) is transmitted to port O1, while the fraction (1−α) of the light power incident at port I1 is transmitted to port O2. This conventional beamsplitter is known as a reciprocal optic element because light which is input into either of the output ports O1 or O2 will reciprocally be transmitted to the input ports I1 and I2. Specifically, a fraction α of any light power incident at port O1 is transmitted to port I2, while the fraction (1−α) of the light power incident at port O1 is transmitted to port I1. Finally, a fraction α of any light power incident at port O2 is transmitted to port I1, while the fraction (1−α) of the light power incident at O2 is transmitted to port I2.
Reciprocal Power Losses in Conventional OCDR and OCT
A second clear drawback of the use of the conventional Michelson interferometer topology in OCDR and OCT is that significant reflected sample arm power is lost because it is inevitably directed back into the source, rather than being collected by the detector 22. In the theoretical analysis which leads to Eq. (1) (derived in the limit of a low reflectivity sample) the noise power is proportional to the amount of power incident on the detector 22 from the reference arm 18, while the signal power is proportional to the product of the coupler splitting ratios from the source 16 to the sample arm 12 and from the sample arm 12 to the detector 22. In the 50/50 (α=0.5) Michelson interferometer used in conventional OCT (see FIG. 1), the light from the broadband source 16 is split evenly between the sample and reference arms, while light returning from both the sample and reference arms is split again into the input arms 34 and 36 containing the source 16 and detector 22. Thus, the detected signal power is proportional to the product of the 50% splitting ratio from the source 16 to the sample arm 12, and the 50% splitting ratio from the sample arm 12 to the detector 22, for a combined sample power double splitting ratio of 25%. Fiber couplers with coupling ratios other than 50% are commonly available; however, their use in the Michelson configuration is even worse. For example, if a 90/10 beamsplitter directs 90% of the source light into the sample arm and only 10% of the light into the reference arm, then the combined sample power splitting ratio is only 9% (10% of the 90% of the source light power incident on the sample).
The modified form of Eq. (1) which is correct for the case of arbitrary splitting ratio is set forth as Eq. (2):
                    SNR        =                              ρ            ⁢                                                  ⁢                          P              O                        ⁢                          R              s                        ⁢                          α              ⁡                              (                                  1                  -                  α                                )                                              qB                                    (        2        )            Here, α is the coupler splitting ratio and Po is the source power. Eq. (2) is consistent with Eq. (1) since in the case of Eq. (1), Ps=Po/2. Clearly, the SNR in Eq. (2) is optimized for α=0.5, or a 50% coupling ratio.Motivation for the Invention
Until the development of the present invention, the only method to increase the sensitivity or acquisition rate in OCDR and OCT was to increase the source power. Increases in the source power are very expensive given current source technology. The design of conventional OCDR and OCT interferometers with reciprocal beamsplitters, is very inefficient with the expensive source power, since up to 50% of the source power is lost due to attenuation of the reference arm, and an additional 50% of the power reflected from the sample is wasted by being directed back into the source. An interferometer design which avoids both of these problems could be up to a factor approaching 4 more efficient, and could thus obtain better quality images at the high speeds required for commercial applications of OCDR and OCT technology. Thus, there is a clear need for an invention which makes more efficient use of broadband source light than the conventional OCT interferometer.